On 02/03/2011 11:12 AM, Davidlohr Bueso wrote:
It wouldn't hurt to add some parenthesis in the following two
similar expression for better readability.
I take it you were using some gcc option about precedence of operations
being confusing between ! and <; including that gcc warning message in
your commit message can aid the discussion.
---
src/util/macvtap.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/util/macvtap.c b/src/util/macvtap.c
index 09d7b78..92147ab 100644
--- a/src/util/macvtap.c
+++ b/src/util/macvtap.c
@@ -1048,8 +1048,8 @@ doPortProfileOpSetLink(bool nltarget_kernel,
memcpy(ifla_vf_mac.mac, macaddr, 6);
- if (!nla_put(nl_msg, IFLA_VF_MAC, sizeof(ifla_vf_mac),
- &ifla_vf_mac) < 0)
+ if ((!nla_put(nl_msg, IFLA_VF_MAC, sizeof(ifla_vf_mac),
+ &ifla_vf_mac)) < 0)
The indentation is wrong, too; the &ifla... of the second line should
line up just after nla_put( of the first line. That was botched in
December by commit 013c000 when the lines were reindented.
Actually, you've uncovered a bigger bug, but didn't fix it right.
Here, 'expr' is 'nla_put(...)'. But '!expr < 0' will never be
true.
('!expr' is either 0 or 1, which is always non-negative). We either
meant 'expr < 0' or '!expr', but not both (or maybe we meant
'!(expr <
0)', which is easier if written 'expr >= 0'). The bug has existed since
Stefan's commit 77e7c13 in November; and looking at the rest of that
commit, it _looks_ like we meant to use 'expr < 0' (that is, the correct
fix is to delete !), but I'd like confirmation from Stefan.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org